Prediction of train delay time is the main foundation for formulating operation strategies in the train dispatching command system. When a train cannot arrive on time due to unexpected events, it is necessary to effectively adjust the train operation plan in time to reduce the impact on other trains and ensure the orderly operation of transportation. In accordance with the characteristics of high-speed railway such as higher traffic density and randomness delays, a Grey Markov model is proposed, which a GM (1,1) model is constructed, followed by the Pigeon-Inspired Optimization (PIO) algorithm for parameter optimization. Then the prediction results are corrected by the GM model, and the predicted values are used to dynamically update the raw data, to improve the prediction accuracy. The test results show that the method for prediction of train delay time based on grey Markov is highly accurate and feasible.